a comparative evaluation of wi-fi rtt and gps based ...€¦ · wi-fi-based positioning technology...
TRANSCRIPT
International Global Navigation Satellite Systems Association IGNSS Symposium 2020
Colombo Theatres, Kensington Campus, UNSW Australia
5 – 7 February 2020
A Comparative Evaluation of Wi-Fi RTT and GPS Based Positioning
Yuntian Brian Bai (1) School of Science, RMIT University, Australia
Allison Kealy (2) School of Science, RMIT University, Australia
Guenther Retscher (3) Dept of Geodesy and Geoinformation, Vienna University of Technology, Austria
Lucas Hoden (4) School of Science, RMIT University, Australia
ABSTRACT
Wi-Fi-based positioning technology has snowballed over the past 20 years
along with the fast development and applications of smartphones for indoor
positioning. On the other hand, Wi-Fi is increasingly accepted for outdoor
positioning due to the availability and popularity of public Wi-Fi in global
cities. Since GPS signals are often interrupted and unstable in the downtown
areas with high-rise surrounded, Wi-Fi becomes an ideal positioning
technology as a substitution of GPS. Especially after the release of the IEEE
802.11mc standard last year, researchers and specialists from industries were
attracted immediately after the release. The new standard provides a fine time
measurement protocol for us to use multiple round-trip time (RTT) rather than
the received signal strength indicator (RSSI) for calculating the distance
between a Wi-Fi access point (AP) and a mobile end-user device. This paper
presents an evaluation and comparison study between Wi-Fi RTT and GPS
based localisations in an outdoor space located in a central downtown area in
Melbourne city. Based on the same testing environment and the same testing
points within a central city area, both GPS and Wi-Fi RTT are tested and
analysed. Results showed that the average positioning accuracies from the two
technologies are 5.10 m and 1.40 m, respectively. The Wi-Fi RTT technology
demonstrated a much better performance both in accuracy and stability.
KEYWORDS: GPS, Wi-Fi RTT, positioning, smartphone, LBS.
1. INTRODUCTION
The recent release of the IEEE 802.11mc standard provides us with a new era for smartphone
and Wi-Fi-based localisation. As the single most popular wireless network protocol of the 21st
century, Wi-Fi technology powers not only most home and business wireless networks, but also
public hotspot networks (Mitchell 2019; Ta 2018). Wi-Fi and smartphone-related location-
based service (LBS) and indoor positioning have gained much attention from both research and
industrial communities in the recent ten years (Donovan 2013; Machowinski 2013; Elkhodr,
Shahrestani, and Cheung 2016; Mohapatra, Choudhury, and Das 2014; Adams 2018; Gao,
Tang, and Bai 2019; Bai et al. 2014; Bai 2016).
The advantage of the 802.11mc standard for localisation is that supports a fine-time-
measurement (FTM) protocol, which allows us to calculate the distance between a smartphone
and an AP using round-trip-time (RTT) of the Wi-Fi signal transmission between a smartphone
and an AP. Applying the Wi-Fi RTT protocol leads to the increment of the positioning accuracy
from 5-10 meters as obtained from traditional positioning methods to about 1 meter in any line-
of-sight (LoS) surrounding environment (Diggelen, Want, and Wang 2018). This has brought
us a great era in using smartphone-based LBS, as both hardware standard and Android
application programming interfaces (APIs) are simultaneously evolving to enable an improved
ranging accuracy that has not previously been possible when using smartphones and Wi-Fi.
The rest of the paper is outlined as follows: the principle of the Wi-Fi RTT protocol will be
introduced in Section II. Section III will discuss the combination and conversions between
geodetic and local coordinate systems and also the Wi-Fi-based multilateration process. Section
IV will present the procedure of experimental tests and results analysis. Finally, the conclusion
part and future work will be addressed in Section V.
2. HOW DOES WI-FI RTT WORKS
The Wi-Fi 802.11 standard provides a possible way of achieving high-accuracy positioning in
a dense multipath environment, which imposes several hardware design changes in the existing
WLAN chipsets in order to increase the timing resolution from the microseconds level to the
nanosecond level (or even sub-nanosecond level) (Diggelen, Want, and Wang 2018). The Wi-
Fi RTT is a point-to-point (P2P) single-user protocol, which includes an exchange of multiple
message frames between an initiating station (ISTA) and a responding station (RSTA). The
ISTA (e.g., a smartphone) attempts to measure its range to the RSTA (e.g., an AP). Obtaining
an accurate time-delay estimate in a dense-multipath environment is challenging. It requires
precise detection of the first signal path associated with the LoS condition between the two
stations and the estimation of its arrival time (Banin et al. 2017; Yu et al. 2019). That is why
the RTT protocol not entirely compatible with a non-line-of-sight (NLoS) surrounding
environment currently. While Wi-Fi RTT protocol enables distance ranging between a
smartphone and an AP, the whole procedure is described as follows. First, the smartphone sends
an FTM request to the AP, then, the AP receives the request and returns an acknowledgement
(ACK) signal to the phone terminal. After that, several FTM feedbacks are sent from the AP to
the mobile terminal, and, then, the mean round-trip time is used for range calculation. This
process can also be performed between several smartphones and Wi-Fi APs at the “same” time.
The whole FTM RTT procedure is shown in Figure 1, in which the number of RTT (also called
“burst number”) can be changed to improve the FTM accuracy by providing multiple
measurements within one period. Currently, the default number of RTT is 8 and the maximum
number of successful measurements is 7.
Figure 1. Principle of the FTM protocol
The mean RTT of each period is calculated by Equation (1) (Yu et al. 2019):
𝑡𝑅𝑇𝑇 = 1
𝑁 (∑ 𝑡4_𝑖
𝑁
𝑖=1− ∑ 𝑡1_𝑖
𝑁
𝑖=1) −
1
𝑁 (∑ 𝑡3_𝑖
𝑁
𝑖=1− ∑ 𝑡2_𝑖
𝑁
𝑖=1) (1)
where:
𝑡1_𝑖 is the timestamp when the FTM framework first sent by a Wi-Fi AP;
𝑡2_𝑖 is the timestamp when the FTM signal arrives at the smartphone;
𝑡3_𝑖 is the timestamp when the smartphone returns the ACK signal to the AP;
𝑡4_𝑖 is the timestamp when the ACK signal is finally received by the AP;
𝑁 is the successful burst number (where 𝑁 > 0, 𝑁 < 𝐵 ); and
B is the total burst number (i.e., burst size, B = 8 by default in this research).
Generally, the protocol excludes the processing time on the smartphone terminal by subtracting
(t_(3_i) - t_(2_i)) from the total round-trip time (t_(4_i) - t_(1_i)), which represents the time
from the instant the FTM message is sent (t_(1_i)) to the instant, the ACK is received (t_(4_i)).
This calculation is repeated for each FTM-ACK exchange, and the final RTT is the average
over the successful number of FTM-ACK per burst. The estimated range can be obtained
through Equation (2).
Estimated Distance: Dest =1
2∗ 𝑡𝑅𝑇𝑇 ∗ 𝑐 (2)
The precondition for a smartphone to support Wi-Fi RTT is that the Android Pie (or called
Android P) operating system (OS) installed on it, which provides a number of APIs and allows
a developer to add RTT methods in their own application. The application needs to declare the
ACCESS_FINE_LOCATION permission, and both location and Wi-Fi scanning need to be
enabled on the end-user device [14].
One disadvantage of the Wi-Fi RTT so far is that it is hard to find many devices fully supporting
the RTT protocol. The RTT-based ranging requires supports from both the ISTA and RSTA
sides, which means all the devices must implement the 802.11mc standard (Android 2018).
Although many smartphone manufacturers announced that their products support the RTT
standard, e.g., the Essential, Nokia, OPPO, VIVO, Sony, and Xiaomi smartphones, none of
them (except the Pixel phones) has successfully tested and recorded by researchers as a device
supporting the Wi-Fi RTT protocol. From a simple test conducted by us, both OPPO Reno 5G
and VIVO X27 are not supporting the Wi-Fi RTT (see Figure 2) and only the Pixel 3 phone
supports the new protocol. For the RSTA devices, only the CompuLab WILD Wi-Fi RTT router
was formally announced so far to support the Wi-Fi RTT protocol. However, a few Google APs
were also successfully configured by us to do so.
Figure 2. Results of Wi-Fi RTT supporting test from three different smartphones
3. METHODOLOGY 3.1 Conversions between different coordinate systems
A universal geographic coordinate system needs to be established for both positioning systems
in order to compare the positioning accuracy between GPS and Wi-Fi RTT. Firstly, the east,
north, up (ENU) Cartesian coordinate system is defined according to the vertical and horizontal
dimensions from the latitudes and longitudes received, earth-fixed (ECEF, also known as earth-
centred rotational (ECR)) coordinate system. The ENU coordinates are formed from a plane
tangent to the Earth's surface fixed to a specific location and hence it is sometimes known as a
local tangent or local geodetic plane (see Figure 3). The relationship between ENU and ECEF
coordinate systems is shown in Figure 3.
Figure 3. Relationships between the local ENU and the ECEF coordinate systems
A local coordinate system called 𝐸′𝑁′𝑈′ Cartesian coordinate system is also established based
on the 𝐸𝑁𝑈 coordinate system, and the only difference is that the 𝐸′𝑁′𝑈′system reset the
values of E, N and U to 0 as a new initial point of the 𝐸′𝑁′𝑈′ system. Then, another local
coordinate system is also established as the Wi-Fi localisation coordinate system. The Wi-Fi
coordinate system (presented as X, Y and Z) is set to the same initial point with the
𝐸′𝑁′𝑈′system but with an angle (θ) as shown in Figure 4. In this case, let 𝑍 = 𝑈′, so only 2D
coordinate systems are displayed.
Figure 4. Relationship between the local 𝐸′𝑁′ and XY 2D coordinate systems
Generally, the coordinates received by smartphones are geodetic coordinates. The comparison
of the GPS estimates and Wi-Fi estimates require to convert the geodetic coordinates to 𝐸𝑁𝑈
coordinates, then to the 𝐸′𝑁′𝑈′ coordinates, which is usually in a three-stage process:
1. Convert geodetic coordinates to 𝐸𝐶𝐸𝐹 coordinates
2. Convert 𝐸𝐶𝐸𝐹 coordinates to 𝐸𝑁𝑈 coordinates
3. Convert 𝐸𝑁𝑈 coordinates to 𝐸′𝑁′𝑈′coordinates
The above conversion processes can be summarized as:
(𝜙, 𝜆, ℎ) => (𝑋𝑒𝑐𝑒𝑓, 𝑌𝑒𝑐𝑒𝑓 , 𝑍𝑒𝑐𝑒𝑓)
=> (𝐸, 𝑁, 𝑈)
=> (𝐸′, 𝑁′, 𝑈′)
In summary, a geodetic coordinate received from GPS is converted to a local coordinate system.
Equation (04) is used for the conversion between 𝐸𝑁𝑈 and 𝐸′𝑁′𝑈′ coordinates.
{𝐸 = 𝐸′ + 𝐸0
𝑁 = 𝑁′ + 𝑁0 (3)
Accuracies can be compared between [𝐸′
𝑁′]s from the GPS and the positions from local Wi-Fi
RTT.
3.2 Wi-Fi RTT-based positioning processes
Multilateration and simplified least squares (LS), for estimating the position from Wi-Fi RTT.
The estimated distances from smartphones need to be calibrated using an experimental
correction value, which can be obtained from an initial evaluation test.
Four APs with the strongest values of received signal strength indicator (RSSI) is selected for
conducting the LS process if there are more than 4 APs connected.
Compared to the latitudes and longitudes received, the height values form GPS is less accurate;
on the other hand, the height of the APs usually are constant values. Therefore, only 2D
coordinates are concerned in this research. All 3D distance values are simplified to 2D values,
which means only X and Y coordinates are considered for the multilateration process.
Assuming P(x, y) is the target position of a TP to be estimated. The exact formulas are:
𝑑𝑖2 = (𝑥 − 𝑥i)2 + (𝑦 − 𝑦i)2 (4)
or: 𝑥2 + 𝑦2 − 2𝑥i 𝑥 − 2𝑦i 𝑦 = 𝑑𝑖2 − 𝑥𝑖
2 − 𝑦𝑖2 (5)
Let 𝑝 = 𝑥2 + 𝑦2, and
𝑋 = [𝑝 𝑥 𝑦]𝑇
B = [1 −2𝑥1 −2𝑦1
⋮ ⋮ ⋮1 −2𝑥𝑛 −2𝑦𝑛
]
L = [𝑑1
2 − 𝑥12 − 𝑦1
2
⋮𝑑𝑛
2 − 𝑥𝑛2 − 𝑦𝑛
2]
where: 𝑖 = 1, 2, … , 𝑛 and 𝑛 = 3 𝑜𝑟 4. 𝑋 can be calculated by:
𝑋 = (𝐵𝑇𝐵)−1 𝐵𝑇 𝐿 (6)
Finally, the coordinates of the point P(x, y) can be obtained from Equation (12).
4. TEST & ANALYSIS 4.1Testbed establishment
As shown in Figure 5a, an area of about 30 x 25 𝑚2, located in the RMIT Alumni Courtyard
behind the Old Gaol, was selected as the testing area. Twelve testing points (TPs) were marked
on the ground, and the distances in X and Y directions to each adjacent points are all 6 metres
as shown in Figure 5b. The 4 APs are placed in the same height level with the smartphone so
that the height values for both the APs and the smartphone were omitted during the
multilateration process. The coordinates of the 4 APs in the local XY coordinates are shown in
Table 1.
Table 1. Coordinates of the 4 APs
(a) Testing area: RMIT’s Alumni Courtyard behind the Old Gaol
(b) The 12 TPs and 4 APs located in the testing area
Figure 5. Testbed establishment in city campus of the RMIT University
The Leica GS18 receiver (see Figure 6), a high precision GNSS receiver, was firstly used for
determining the coordinates of 4 base points (TPs 1, 3, 10 and 12) as their true coordinates, then
other true coordinates of the 8 TPs and 4 APs were defined accordingly with the assistance of
a total station and a tape.
Figure 6. Definition of the true coordinates for the 12 TPs
A Pixel 3 smartphone was used for collecting the data from GPS and the Wi-Fi APs. Two
Android APPs called GPS coordinates and Wi-Fi RTT Scan (see Figure 7) were used
respectively for collecting the latitude and longitude data and RTT data at each TP. Five pairs
of data were collected at each TP, and the average values were used as the location estimation
process. The time interval between any two adjacent collections is 2 seconds.
Figure 7. APPs used for colleting the GPS and Wi-Fi RTT data
4.2 Process and analysis of the received data
After the latitudes and longitudes of the 4 base TPs were obtained from the Leica GNSS
receiver, the true coordinates of the other 8 TPs can be easily defined by transforming the local
XY coordinates to EN coordinates with known bearing (i.e., θ, see Figure 4). In this research,
θ is defined as -14.437°, calculated from the true coordinates of TP1 and TP12. The received
latitudes and longitudes from the Android APP need to be converted to the grid coordinates for
the accuracy comparison process later. All the data including the true EN coordinates received
latitudes and longitudes, and the EN coordinates converted from the latitudes and longitudes,
as well as the accuracies for the 12 TPs, are listed in Table2.
It is noticeable that all the accuracy values for the first 4 TPs are more than 11.917 m and then
immediately reduced to the level of 1.824 to 7.226 m. The possible reason for this is that the
GPS coordinates APP might not work in a stable mode at the beginning of 1.5 minutes or other
disturbances, for example, the user had to stand by the smartphone for initial operation and the
user’s body might partly block the GPS signals. Without consideration of the first four accuracy
values, an average accuracy of 5.10 m was obtained from the rest 8 TPs for the GPS-based
positioning, which is as the average level as expected. All these accuracy values are displayed
in Table 2 and Figure 8.
Table 2. Process of the data received from GPS
Figure 8. Positioning accuracies based on the data received from GPS
While collecting the data from GPS, Wi-Fi RTT ranging data were also collected in the same
period. For each TP, the four ranging values received from the 4 APs need to be calibrated by
adding a deviation correction value (i.e., -3.348 m) according to the research results from (Bai
and Kealy 2019), then, they were applied to a multilateration process through Equations (10)
to (12). The accuracy value for the TP was obtained based on its estimated local X and Y
coordinates. The results and relevant data are displayed in Table 3 and Figure 9.
Table 3. Process of the data received from the Wi-Fi APs
Figure 9. Positioning accuracies based on the data received from the Wi-Fi APs
There are still some issues that need to be considered for the above positioning estimation
process. Apart from the unstable mode of the smartphone at the beginning of 1.5 minutes, other
factors also exist to affect the positioning accuracies, such as multi-path effect, effects from the
people walking around in the area during the test, and the smartphone user, who have to stand
by the phone in a very near distance to operate the phone.
The abovementioned test results show that positioning with Wi-Fi RTT can obtain much better
accuracy than with GPS. As Wi-Fi services are more and more common and free available in
many world-wide cities, location-based service (LBS) with Wi-Fi applications will become
more and more popular and vital for people’s daily activities.
3. CONCLUSIONS
This paper presents an evaluation and comparison study of GPS and Wi-Fi RTT based
localisation using a smartphone as the end-user device. The experimental test was conducted in
a relatively open area of the town centre in Melbourne. Relevant data from both GPS and Wi-
Fi RTT are collected simultaneously from the same TPs, and other user environmental factors
for both GPS and Wi-Fi are also the same. Results showed that the average accuracy for Wi-Fi
RTT based localisation is 1.40 m and much better than the accuracy obtained from the GPS.
This result is also firmly promising compared with the traditional way of Wi-Fi and
smartphone-based localisation results.
Implementation and comparison of these technologies in a real-time kinematic mode will
become the next step of our research plan.
REFERENCES
Adams, Dana. 2018. "WiFi Planning: Preparing for Growth in a Mobile-First World." In.: SecurEdge
Networks.
Android. 2019. "Android Documentation." Google Developers, Accessed 9 Sept.
https://developer.android.com/guide/topics/connectivity/wifi-rtt.
Bai, Yuntian Brian. 2016. "Development of a Wi-Fi and RFID Based Indoor Location and Mobility
Tracking System." PhD Thesis, RMIT University.
Bai, Yuntian Brian, and Allison Kealy. 2019. "Evaluation and Correction of Smartphone-Based Fine
Time Range Measurements (Submitted)." In The Sixth International Conference on Ubiquitous
Positioning, Indoor Navigation and Location-Based Services (UPINLBS). Beijing, China.
Bai, Yuntian Brian, Suqin Wu, Guenther Retscher, Allison Kealy, Lucas Holden, Martin Tomko,
Aekarin Borriak, Bin Hu, Hong Ren Wu, and Kefei Zhang. 2014. "A new method for improving
Wi-Fi-based indoor positioning accuracy." Journal of Location Based Services 8 (3):135-47.
Banin, Leor, Ofer Bar-Shalom, Nir Dvorecki, and Yuval Amizur. 2017. "High-Accuracy Indoor
Geolocation using Collaborative Time of Arrival." In. Petah Tikva 49527: Intel Corporation.
Diggelen, Frank Van, Roy Want, and Wei Wang. 2018. "How to achieve 1-meter accuracy in Android."
In GPS World. 1360 East 9th Street, Suite 1070, Cleveland, OH 44114: North Coast Media.
Donovan, Fred. 2013. "Wireless LAN Market to Expand 57 Percent by 2017." In Networks and
infrastructure, edited by Dell'Oro Group. FierceMobileIT.
Elkhodr, Mahmoud, Seyed Shahrestani, and Hon Cheung. 2016. "Emerging Wireless Technologies in
the Internet of Things: a Comparative Study." Review of. International Journal of Wireless &
Mobile Networks (IJWMN) 8 (5):67-82. doi: 10.5121/ijwmn.2016.8505.
Gao, Longshi, Shuangxia Tang, and Yuntian B. Bai. 2019. "A New Approach for Wi-Fi-Based People
Localization in a Long Narrow Space." Review of. Wireless Communications and Mobile
Computing 2019 (3):1-8.
Machowinski, Matthias. 2013. "New high of $4 billion for wireless LAN market in 2012." In Wireless
LAN Equipment and WiFi Phones. Infonetics Research.
Mitchell, Bradley. 2019. "Introduction to Wi-Fi Wireless Networking." In, 1-3. 1500 Broadway, New
York, NY 10036: Lifewire (part of the Dotdash).
Mohapatra, Sumant Ku, Ramya Ranjan Choudhury, and Pravanjan Das. 2014. "The Future Directions
in Evolving Wi-Fi: Technologies, Applications and Services." Review of. International Journal
of Next-Generation Networks (IJNGN) 6 (3):13-22. doi: 10.5121/ijngn.2014.6302.
Ta, Viet-Cuong. 2018. "Smartphone-based indoor positioning using Wi-Fi, inertial sensors and
Bluetooth." Université Grenoble Alpes.
Yu, Yue, Ruizhi Chen, Liang Chen, Guangyi Guo, Feng Ye, and Zuoya Liu. 2019. "A Robust Dead
Reckoning Algorithm Based on Wi-Fi FTM and Multiple Sensors." Review of. Remote Sensing
11 (5):1-22.